Methodology

We look for documents, public datasets, official records and reports from recognized outlets. Each source must be opened and read: a search title or summary is not enough to support a claim.

We check what each figure measures, the period it refers to and the publication date. References are linked to claims and collected at the end of each page, with links for readers to consult. Information we cannot verify is left out.

Research determines what the page can claim. The headline is adjusted when the documents do not support the initial idea. We distinguish a person’s actions from those of their party, allies or a government.

Investigation, formal accusation and conviction are different situations. We record the status described by the source, include the defense of those involved where relevant, and consider counterpoints even when they contradict the project’s position.

We only place values on the same scale when they measure comparable things. A budget amount is not equivalent to proven diversion of funds. Overlapping values are not added together, and each comparison states what is being measured and over which period.

When comparing monetary values from different periods, we adjust for inflation using the IPCA in the Central Bank’s series and state the reference month. We check the calculations and use conservative criteria: ratios presented as “more than” are rounded down.

We organize each topic into a headline, a visualization when it helps explain the data, a short text and references. The argument’s strength must come from facts, without insults or speculation. Each claim is checked against its supporting source, and the content is reviewed by the project’s contributors.

We also check readability on phones and computers, links and accessibility. The work remains open to corrections: if you find an error, send the page, passage and a source for verification through the channels on the page about the project.

We record how often each page is opened, read to the end and shared, to learn which topics attract the most interest. Alongside this, only categories: where the visit came from (WhatsApp, Instagram, search), device type, country and, in Brazil, state.

For outreach material, we count how often each piece is downloaded and its QR code scanned: the QR code links to the site with the piece’s name and format, which we remove from the address as soon as the page opens. Nothing identifies who downloaded, printed or read it.

We do not use cookies, identify readers or store IP addresses. If your browser requests not to be tracked, we record nothing.